Liquid AI Launches LFM3 Multimodal Foundation Model

All models are 2B-parameter and designed for on-device fine-tuning and deployment. LFM3-1.7B-Base features a tri-hybrid architecture with improved parameter efficiency and long-context performance. Vision and audio models include new in-house encoders that reduce compute by up to 3x while maintaining or improving accuracy.

Summarised from the original article by an open-source model, then checked automatically: every number and English name must appear in the source. The original article is authoritative.